A PCB stackup is the recipe for how your board is built up in layers: the copper layers where you route, and the insulating material between them. It sounds like a fab detail, but it quietly decides how well your board handles signals, power, and noise — long before you place a single component.
Short answer: The stackup is the ordered set of copper and dielectric layers in your board. Two layers work for simple, low-speed designs. Four layers — with dedicated ground and power planes — give you far better signal integrity, cleaner power, lower EMI, and predictable controlled impedance. Most non-trivial digital boards belong on four layers.
What a PCB stackup actually is
A rigid PCB is a sandwich. The layers you care about are:
- Copper layers — the conductive sheets where traces, pads, and planes live. A "2-layer" board has copper on top and bottom; a "4-layer" board adds two internal copper layers.
- Core — a cured piece of fiberglass-and-resin (usually FR-4) with copper bonded to one or both sides. It is rigid and forms the structural backbone.
- Prepreg — "pre-impregnated" fiberglass that is still partly uncured. During lamination it melts and bonds the layers together, then hardens.
- Dielectric — the general term for that insulating material between copper layers. Its thickness and dielectric constant (Dk) set the capacitance between layers and directly affect trace impedance.
So a stackup isn't just "how many layers." It's the specific order, the thickness of each dielectric, and the material properties — all of which the fab needs to build the board and which your design has to match.
2-layer vs 4-layer: what changes
On a 2-layer board, both copper layers usually do double duty: routing and whatever ground and power you can fit. That's fine for a power supply, a simple sensor board, or a low-speed microcontroller project. It's the cheapest option and every fab makes them.
The trouble starts when the board gets busy. With only two layers, your ground return path is wherever a trace happens to go, power is a web of traces instead of a plane, and there's simply not enough room to route a dense design without compromises.
A 4-layer board adds two internal layers, and the usual move is to dedicate them to planes. A typical arrangement looks like this, top to bottom:
- Signal (top, components + routing)
- Ground plane (solid copper)
- Power plane (solid copper)
- Signal (bottom, routing)
That single change — putting solid ground and power planes in the middle — is the reason most engineers reach for four layers.
Why 4 layers helps signal integrity, power, and EMC
Three things improve at once when you add those internal planes:
Signal integrity. Every signal layer now sits directly above a solid reference plane. High-speed return current flows on that plane right underneath the trace, tightly coupled, instead of wandering across the board. That means fewer impedance discontinuities, less crosstalk, and cleaner edges.
Power distribution. A dedicated power plane paired with the ground plane behaves like a distributed capacitor, giving your ICs a low-impedance supply and a place for decoupling to work. No more routing power as a fragile daisy chain of traces.
EMC / EMI. A continuous ground plane is the single most effective thing you can do to reduce radiated emissions. It shrinks current loop areas and gives noise a clean, short path back to its source — which is what gets boards through emissions testing.
You don't get any of this "for free" on two layers; you get it by choosing the right stackup.
When you need 6 or more layers
Four layers cover a huge range of real products, but you go higher when:
- You need more routing space than four layers give (dense BGAs, many buses).
- You want multiple reference planes so high-speed signals can be sandwiched between grounds.
- You're routing high-speed interfaces (DDR, high-lane-count SerDes, fast differential pairs) that need tight, well-referenced impedance control.
More layers mean more cost and more manufacturing complexity, so the honest rule is: use the fewest layers that let you route cleanly with proper reference planes.
| Layers | Typical use case | Relative cost | Choose when |
|---|---|---|---|
| 2-layer | Power supplies, simple sensors, low-speed MCU boards | Lowest | Simple, low-speed, plenty of routing room |
| 4-layer | Most IoT, MCU, and mixed-signal products | Moderate | You need solid planes, controlled impedance, or more space |
| 6-layer+ | Dense digital, high-speed buses, multi-plane designs | Highest | Routing/plane needs exceed four layers |
How the stackup drives controlled impedance
Controlled impedance means a trace presents a specific characteristic impedance — commonly 50 Ω single-ended or 90/100 Ω differential — so fast signals don't reflect and distort. That number isn't set by the trace alone. It's a function of trace width, copper thickness, the dielectric height to the reference plane, and the material's Dk.
All three of those last factors come from the stackup. A trace over a plane 0.1 mm away has a very different impedance than the same trace over a plane 0.3 mm away. This is exactly why you can't design controlled-impedance traces on a guess: you need the real layer heights. We go deeper on this in our guide to controlled impedance in KiCad, and impedance targets pair naturally with well-defined net classes and design rules.
Get the real stackup from your fab
The single most useful habit: use your manufacturer's actual stackup, not the tool's defaults. Fabs publish standard 2, 4, and 6-layer stackups with real dielectric thicknesses, copper weights, and Dk values, and most will give you a custom stackup on request. Enter those exact numbers into KiCad's Board Setup so your impedance calculations and clearances reflect what gets built — not a generic assumption.
Cost trade-offs, honestly
Going from 2 to 4 layers isn't free — it's a real jump in bare-board cost. But it's often cheaper than the alternative: a two-layer board that fails emissions testing, needs a redesign, or eats days of debugging noise problems. For anything beyond simple and slow, four layers usually pays for itself. And once you're routing high-speed signals, the extra layers aren't a luxury — they're the only way to hit your impedance and EMC targets. If you're watching unit cost, our notes on reducing PCB manufacturing cost cover where to save without compromising the stackup.
Where GoForFab fits
Choosing a stackup is a design decision — but proving your board is manufacturable is a workflow one. GoForFab connects to your KiCad project and, on every push, regenerates the full manufacturing package with kicad-cli: Gerbers, drill files, BOM, CPL, DRC/ERC reports, and 3D renders. Your layer stackup, impedance choices, and design rules stay in KiCad where they belong; GoForFab keeps the outputs current and reproducible so you always have a clean set to hand your fab. It's free forever to get started — connect a project and see your first manufacturing package in minutes.